A critical part of autonomous driving is to provide a detailed description of the vehicle surroundings. This example shows how to train a model on an Intel® Gaudi architecture system to perform semantic segmentation on a dataset of street-based driving images. The model can be used to provide information to the autonomous driving subsystem control system, employing an algorithm that associates a label or category with every pixel in an image, and is used to recognize a collection of pixels that form distinct categories. In this case, an autonomous vehicle needs to identify vehicles, pedestrians, traffic signs, pavement, and other road features. In the following example, each item category is represented by a separate color.
This example uses the nnU-Net model for semantic segmentation, a self-adapting framework based on the original U-Net 2D model, which is then optimized on the Intel® Gaudi® software stack. We demonstrate how to train this model from scratch using the Audi Autonomous Driving Dataset (A2D2), which features 41,280 images in 38 categories. The example demonstrates how quick it is to train this model on an Intel® Gaudi® platform and how you can further accelerate it using multiple Intel® Gaudi® accelerators. In this case, eight Intel Gaudi accelerators are used, resulting in training of the model in only two hours.
This autonomous vehicle use case is available on GitHub* in a Jupyter* Notebook. To run this use case, start with one of the two following options:
Amazon EC2* DL1 Instances (based on first-generation Intel Gaudi software):
- An Amazon Web Services (AWS)* account is required. For instructions on starting a DL1 instance, see the quick start guide.
Intel® Developer Cloud Using Intel Gaudi 2 Software
- A user account is required.
For more information, see the installation guide.